How does ArduIMU and IMU in general work?
Something has been somewhat confusing to me. No matter where I look, I cant find a straight explanation as to how autonomous aircraft stabilize themselves. I know the Ardupilot uses thermopile sensors. Do real drone aircraft use thermal horizon sensors? I know gyros are commonplace though. As for the IMU, I realize since its internal and gyro based, there has to be some drift. How is the drift reduced to the point where it can be used to fly an aircraft? It it used in conjunction with other sensors to get more points of reference then mix itself in with programming? Somewhat like being able to auto recalibrate in flight? Or is it kind of like a traditional model aircraft gyro, where it just stabilizes flight against unwanted movement. How much of an improvement in stability can an ArduIMU offer over just thermal sensors? I will not be using them anytime soon, I think they are not yet implemented fully as for the open code, but I am just curious. Are there any other sensors commonly used in UAV type aircraft in general?
I wish I could find a good overall guide or explanation for UAV stabilization.
Replies to This Discussion
-
Permalink Reply by Ryan Beall on -
In short IMU's use complicated attitude estimation algorithms that are pretty math entensive. The general idea is to blend or complement two sensors together to get the best of both worlds. Gyro's drift and accelerometers are noisy but thy can both be used to determine attitude. The accelerometers keep the gyro drift stabalized.
Then once you have a stable attitude solution, you use that to keep the wings level using a negative feedback loop of the error. IE: if you want to keep the wings level and they are not then move the ailerons to correct in the opposite direction
-
Permalink Reply by Scott Lencl on -
So it is reliable enough to know its proper orientation without any outside information?
-
Permalink Reply by Scott Lencl on
-
Oh, I think I get it now. So even if the gyro were to drift a bit, and the plane thought it was level, it could use the accelerometers to detect that its turning in another direction, so it would use the feedback from the accelerometers to know its turning and mix the data with the gyro.
-
Permalink Reply by Scott Lencl on
-
the way I am thinking of it, is it similar in a sense to how a segway. What I think of it as is how a segway can detect that its tilting forward and so it simply counteracts that until its not moving anymore. So similarly the plane will balance itself until it detects no more unwanted roll or pitch. I wont be surprised if I am completely wrong, but I just want to get a grasp of how this really works for sure.
-
Permalink Reply by Sgt Ric on -
As Ryan mentioned, its the combination/overlap of the accellerometers and the gyros that work to create the balance needed.
Check out the gyro vs accellerometer thread in the FAQ section of the front page...it's been a hot topic with alot of input from many!
-
Permalink Reply by Scott Lencl on
-
Thanks, perfect!
-
Permalink Reply by Raymundo Magana on -
Seqway sensors are way more expensive... and Drone IMU's waaaaaay more expensive...
as an example, the MTi from Xsens costs about $5,000 and even that one is probably not good enough from a military drone... well probably for a small one. The point is that the cheap gyros and accelerometers that DIY users normally use are low performance and thus the attitude estimation is not so accurate. That being said, using filters and redundancy greatly increments the accuracy and stability of the systems with cheap sensors
The actual "control" is generally done using PID controllers (or a sub-controller, PD, PI or P) that stands for Proportional, Integral and Derivative and is the most common automatic controller used in the world (aproximatedly 90% of all the controllers in the world use this kind of control, not just UAV's).
This controller uses the "estimated" angle given by the IMU or thermopiles and calculates the difference between it and the desired angle... the result is processed by the PID (PD,PI or P) using a mathematical equation and an output is then calculated...
Complex systems, like UAV's, have to use multiple controllers to stabilize the plane.
Altought simple PID controllers are know to be a not "optimal" solution for plane control (PIDs assume that the system is linear, and a plane is a non-linear system) they are the simplest to implement and tune.
Some people think that the effort it takes to implement more complex controllers is normally not seen on the actual performance of the system...
-
Permalink Reply by Marc Vanderbilt on -
I thought the MTi was cheaper, around 2500 USD. They also have a unit with GPS which should be around 5000 USD.
http://damien.douxchamps.net/research/imu
-
Permalink Reply by Raymundo Magana on
-
The price may be cheaper now... when I was intrested in buying one for my thesis the price was arround that... 1.5 or 2 years ago!
I could also be wrong :D... they're expensive non the less
-
Permalink Reply by Scott Lencl on
-
Although fairly inaccurate compared to high end IMUs, If I am correct, an IMU for our model aircraft application should work great for $100-200. Anyone have any commone on this?
-
Permalink Reply by Raymundo Magana on
-
It depends, if you want to have GPS it's a little more expensive than just 9 DoM. Also... for $125 you could get the sensors and an MCU (no GPS) but the code of the comercial products is generaly better tested and debugged... thus it SHOULD have better performance...
You can make it work with the IMU's on that range... but they're trickier to use (vibration!!!)
I love working from scratch so I bought a razor 6 DoF, a LS20032 and a HMC5843... I'm using a dsPIC30F4013 to make the IMU... it should be a LOT of fun! :P
-
Permalink Reply by Scott Lencl on
-
Can the arduimu be used currently with the ardupilot, or is there still code that needs to be written for it.
Top Content
Groups
-
ArduCopter User Group
478 members
-
ArduPlane User Group
299 members
-
ArduCopter (traditional …
77 members
-
ArduPilot (legacy) User …
77 members
-
Moderators
50 members
-
ROS User Group
31 members
-
Team Prometheus
11 members
© 2012 Created by Chris Anderson.
